Diffusive Synchrotron Radiation from Relativistic Shocks of Gamma-Ray Burst Sources

Diffusive Synchrotron Radiation from Relativistic Shocks in

Gamma-Ray Burst Sources

Gregory D. Fleishman 1
1 National Radio Astronomy Observatory, Charlottesville, VA 22903

ABSTRACT

The spectrum of electromagnetic emission generated by relativistic electrons
scattered on small-scale random magnetic fields, implied by current models of
the magnetic field generation in the gamma-ray burst sources, is considered. The
theory developed includes both perturbative and non-perturbative versions and,
therefore, suggests a general treatment of the radiation in arbitrary small-scale
random field. It is shown that a general treatment of the random nature of the
small-scale magnetic field, as well as angular diffusion of the electrons due to
multiple scattering by magnetic inhomogeneities (i.e., non-perturbative effects),
give rise to a radiation spectrum that differs significantly from so-called "jitter"
spectrum. The spectrum of diffusive synchrotron radiation seems to be consistent
with the low energy spectral index distribution of the gamma-ray bursts.

Subject headings: relativity — turbulence — acceleration of particles — gamma
rays: bursts — radiation mechanisms: nonthermal

1. Introduction

Relativistic objects including, but not limited to, Gamma-Ray Burst (GRB) sources
are known to produce nonthermal emission, which typically requires the generation of extra
magnetic field, at least at some stage of ejecta expansion after initial prompt energy release
(Meszaros 2002). A number of recent publications (Kazimura et al. 1998; Medvedev & Loeb
1999; Nishikawa et al. 2003; Jaroshek et al. 2004, 2005) put forward the concept of magnetic
field generation by a relativistic version of the Weibel instability (Weibel 1959). Analytical
considerations (Medvedev & Loeb 1999), together with more sophisticated numerical mod-
elling (Nishikawa et al. 2003, 2005; Jaroshek et al. 2004, 2005; Hededal & Nishikawa 2005),
suggest that the magnetic field produced by the Weibel instability is random and extremely
small-scale, with a characteristic correlation length l cor ~ (0.1 — l)l s k, where l s k = c/u pe is
the plasma skin depth, u pe is the electron plasma frequency, and c is the speed of light.

- 2 -

Depending on the magnetic field saturation value B, this correlation length may be less
than the coherence length (or formation zone) of synchrotron radiation:

Here, e and m are the electron charge and mass, and use — — is the electron gyrofrequency.
Generation of electromagnetic emission by relativistic electrons moving in small-scale mag-
netic fields is known to differ from the case of emission by electrons in a uniform (large-scale)
magnetic field (Landau & Lifshitz 1971).

Electromagnetic radiation in small-scale random magnetic fields was first calculated by
Nikolaev & Tsytovich (1979) using perturbation theory, where the trajectory of the particle
is assumed to be rectilinear, while the acceleration produced by the random variations of the
Lorentz force is taken into account by the first non-zero order of the perturbation theory.
The full non-perturbative theory was later worked out by Toptygin & Fleishman (1987a);
Toptygin et al. (1987), allowing in particular net deflections of the radiating particles that
are not necessarily small (e.g., by a large-scale magnetic field and/or multiple scattering on
small-scale inhomogeneities).

Recently, interest in this emission process in the context of gamma-ray bursts was revived
by Medvedev (2000), who estimated radiation spectrum in the presence of the small-scale
magnetic field and referred to the resulting radiation as "jitter radiation". An oversimplifica-
tion of the estimates performed by Medvedev (2000) resulted in the low- frequency asymptotic
limit of the jitter spectrum, P{oo) oc u>, which is applicable to a simpler (undulator-like) ver-
sion of what is considered here, while may noticeably deviate from the correct spectral shape
in a general case. In this paper a general treatment of the radiation produced by relativistic
electrons moving through small-scale fluctuating magnetic fields, probably available at the
relativistic shocks, is presented. For the completeness and convenience of further applica-
tions, we first formulate a simplified perturbative version of the theory, and then specify its
region of applicability by comparing with the available non-perturbative version. Possible
implications of the general theory for GRB physics are discussed.

eB

c

2. Perturbation theory of diffusive synchrotron radiation

Let us calculate the radiation by a single relativistic particle moving in a random small-
scale magnetic field based on the equation for the radiated energy resulting from perturbation
theory (Landau & Lifshitz 1971)

- 3-

Here 7*

dW w
dw

7

27re 2

1/27, 2

— ) l w cy±l
a;' '

+

ur

2^

dl-

a;

(2)

E/mc 2 in the vacuum, while 7^

7 2 + ^ ) in the presence of the

background plasma, uj' = f(7 -2 + 2 + u) 2 e /u) 2 ), 6> is the angle between the electron speed and
the direction of the emission, w w /j_ is the Fourier transform of the particle acceleration wj_ =
F/ (7717) transverse to the particle speed, and F is the corresponding transverse component of
the Lorentz force. Evidently, for a random Lorentz force the acceleration is a random function
as well, although the average of |w w /j_| 2 includes a regular (not random) component, which
eventually specifies the emission pattern.

To find the averaged value of |w w /j_| 2 , assume that the random Lorentz force (F(r,t),
which can include either magnetic and/or electric fields) perturbing the rectilinear particle
motion, can be expressed as a Fourier integral over uj and k:

F(r,f)

e- i(w *- fa) F wk< fc;dk.

(3)

This force represents a global field, which can vary in space and time in general. However,
the particle acceleration is produced by a local value of this field related to instantaneous
positions of the particle along its actual trajectory. Within the perturbation theory we can
adopt r = r + vt for the independent argument of the force acting on the moving particle:

F(r„ + vf,t) = J e-^-^-^F^dcudk.

(4)

Then the Fourier component F^, specifying the magnitude w u / = F LU //m / ~f, can be found by
direct temporal Fourier transform of (4):

F . ■ = / ^-e^''F(r + vt, t) = \ dudk5(uj' -00 + kv)F^ k e ikr °.

(5)

Now it is straightforward to write down the square of the force modulus | F u ' \ 2 -

I F w , | 2 = J dudkd^dk^-^S^' -lu + kv)6(w' + kiv)F W)k F^ likl . (6)

This value is still fluctuating in the random field, although its mean value is non-zero
(in contrast to the fluctuating field itself F w k ) which is a complex random function with zero
mean). A steady-state level of radiation is evidently specified by the average value (| F u i | 2 )
of (6). The easiest way to perform the averaging is to average Eq. (6) over all possible

-4-

initial positions r of the particle inside the source. In practice, this corresponds to a uni-
form distribution of radiating particles within a volume containing the field inhomogeneities,
which have a statistically uniform distribution within the same volume. Since the exponent
exp(i(k — ki)r ) oscillates strongly for k 7^ ki, we obtain:

iy e ^-^ dro= ^( k _ kl)> ( 7)

where V is the source volume, and, for (| F w / | 2 ) obtain accordingly

(I | 2 > = ^ J dudk5{J -co + kv) I F^, k | 2 . (8)
Then the mean value of the Fourier transform of the particle acceleration is expressed as

, „ ,2 W

J dudkS(u' - lu + kv) I F Wjk | 2 , (9)

m 2 '-f 2 V

which provides a unique correspondence between the temporal and spatial Fourier transform
of the Lorentz force (right-hand side) and the temporal Fourier transform of the particle
acceleration (left-hand side). Substituting (9) into (2), using dummy variables (qo,q) for
(u, k), we obtain finally for the radiated energy per unit range of uj

dW^ (27r) 4 e 2 f°° Juj'\(uj\' ,

duj m 2 c 3/ y 2 V A/272 \w / \tu'

2

dq dq6(uj'-q +qy) | F 90;q | 2 .

(10)

The spectrum of the radiation described by Eq. (10) depends on the statistical prop-
erties of the random force. To be more specific let us consider a random magnetic field,
i.e., I F 90)q | 2 = e 2 | | 2 = e 2 (5 a p — v a vp/v 2 )Bg Q B^* >tl , and introduce the second-order

correlation tensor of the statistically uniform random magnetic field as follows (Toptygin
1985):

K<$(t,t) = (B st!a (R,t)B st ,p(R + r,t + T)) = 7^J dtdRB^R, t)B st ^(R + r, t + r).
Then, express | B^ q | 2 via the Fourier spectrum K^(q ,q) of this correlation tensor 1 :

indeed, B«, q < >q = J dt '%^ dr " exp[i(q t' - qr') - i(q t" - qr")]B st , a (r' , t')B st ^" X) =
/ *^exp[i( g0 T - qr)] / dt'dr' B st , a (v> , t')B st , (i (v> + r,t' + r)) = ^ J exp^r - qr)]K^j (r, r) =
(27r) 4 ^iff (gOiO)- Here we used the correlation tensor denned by Eq. (11), whose Fourier transform is
q) = / ^e^-^K^iv^). Then, multiplying this by {5 a[i - v a v f3 /v 2 ), obtain (12).

- 5 -

I Bi„ q l 2 = ^ («* - v -f) kSm = (12)

where i^(g ,q) = ^?g(9ojq) — is the spectrum of the random magnetic field

transverse to the particle velocity. If the random field is composed of random waves with
the dispersion relation q = qo(q) (in the limiting case of a static random field we have
g (q) = 0), the spectrum takes the form K(q , q) = K(q)5(q — qo(q)).

Substituting all required values into (9) and dividing by the total duration of emission
T we arrive at the radiation intensity (energy emitted per unit frequency per unit time):

J dq dqS(u;'-qo+qy)K(q)S(qo-qo(q L )),

(13)

which eventually results in radiation spectra different from one estimated by Medvedev
(2000).

dL

dto m 2 c 3 7 2

d

00

+

id

3. Illustrative examples

It is now straightforward to consider a few different models of the random magnetic
field. Let us proceed with the case of quasi-static magnetic inhomogeneities, K^l(q ,q) =
K-aB (q)^(qo) j which is applicable for magnetic inhomogeneities moving slower than the radi-
ating particle. In the general case we assume:

<)(q) = ^(3 s 2 t >/(q),

(14)

where the tensor structure of C a p should be consistent with Maxwell's Equation V • B =
and provide the correct normalization J Kal (q)dq = (-B 2 t ).

Consider the isotropic case K^I(q) oc (8 a p — q a qa/q 2 ) f(\ q |) with a specific power-law
for the spectrum of random magnetic field:

/(q)

(ql + qiyl^

(15)

where v is the high-frequency spectral index of the magnetic turbulence. The integral over
<iq then takes the form:

dq

q 2 5(uu' + qv) _ 2tt
{ql + g2)^/2+2 - ~

-i//2

-v/2-1

(16)

- 6-

Subsequent integration over d(u' /uj) yields the radiation spectrum over the entire spec-
tral range. We consider the low-frequency and high-frequency asymptotic behavior of the
spectrum. At high frequencies, uj cj 7 2 , where c^o = q m c, q m can be discarded everywhere
in Eq. (16), so after integration over d{uj'/uj) we have dl^/du oc uj~ v and the radiation
spectrum resembles the turbulence spectrum at q ^> q m . At low frequencies, uj <C uJoj 2 , the
term uj'/v can be discarded in (16), which leads to the frequency-independent part of the
spectrum dl^/duj oc uj°. At even lower frequencies, uj < uJ pe ^, when the difference between
7 and 7* is important, the term cu'/v dominates again giving rise to the spectral asymptote
dl^/duj oc uj 2 .

Note that this property does not depend on the specific shape of the correlation function
/(q) in the range of small q <C q m (oc q 2 in our case), and remains valid for any shape peaking
at q m . To see this explicitly, consider the following simplified form of the spectrum: /(q) oc q a
at q < q m and /(q) = at q > q m . Therefore, in place of (16) we have:

which does not depend on w at w' < uj (or uj <C uJo^ 2 ) f° r a > ~ 2, thus, for any fluc-
tuation energy spectrum, q 2 f(q), peaking at q m . In particular, this is also valid for the
Bremsstrahlung spectrum, which arises as the fast particle experiences random accelerations
at microscopic scales by randomly distributed Coulomb centers.

We now demonstrate that the same is valid for the two-dimensional geometry implied by
the current models of magnetic field generation at the relativistic shocks (Medvedev & Loeb
1999; Nishikawa et al. 2003; Jaroshek et al. 2004, 2005). Consider a specific case, when
the observer looks along the x axis, the shock front lies in the xy plane, and the random
field belongs to the shock front, first examined by Medvedev (2000), who obtained a result
different from one given below. Note, that now the most general form of K a/ 3(q) is

K al^) « (<W - n * n fi) f(Qz, Ql + (18)

where n is the unit vector normal to the shock front. Since we observe along the x axis,
particles moving along this direction are the main contributors to the emission, so that
Q v — qx v i n the argument of 5-function. Thus, we obtain:

J d^icj' + q x v)f(q) = l -J dq y f(q 2 + (uj'/v) 2 ), (19)

where f{q 2 + qy) — J f(q)dq z . The correlation function /(q) has a peak around q m = 2n/l cor .
Assuming the function / keeps this property, which is typically the case, then at small

a+2

a+2

(17)

7

frequencies (u/ <C cj ) integral (19) does not depend on frequency, and the substitution of it
into (13) gives rise to a frequency-independent spectrum dl w /dui oc u°, which is flatter (not
steeper) than the asymptotic spectrum of synchrotron radiation.

Thus, this flat low-frequency spectrum seems to be rather common in the presence
of small-scale random magnetic or electric fields. However, in some specific models of the
random field, the low-frequency spectrum can deviate from the flat spectrum.

As an example of this latter point, consider a (somewhat artificial) factorized correlation
function /(q) = /i(?||)/2(q±), where q\\ and q^ stand for components parallel and transverse
to the particle velocity (note that a special model of the field, which is uniform in the plane
transverse to the particle velocity, when /(q) = /i(<?||)<Kq±), is absorbed by this particular
case). Then, integrations over dq\\ and <iq_i_ can be performed independently, so that

and the low-frequency asymptote of the radiation spectrum is ultimately specified by the
behavior of fi(q\\) at q\\ <C q m (e.g., fi(u;'/v) oc (uj'/v) 2 for correlation function like (15)), and
can easily be steeper than u° spectrum with the steepest possible asymptote u l (ignoring
non-perturbative effects and wave dispersion).

To see this, consider an extreme case of a small-scale field, namely, one comprising a
single spatial harmonic /(q) = 5(q — q m ), which evidently provides the narrowest possible
radiation spectrum. All integrations are extremely easy in this case, in particular,

giving rise to the linear low- frequency asymptote dl^/du oc u 1 , in agreement with the
estimate of Medvedev (2000). A similar kind of regular (but small-scale) acceleration takes
place in undulators or in the case of so-called small-pitch-angle radiation (Epstein 1973;
Epstein & Petrosian 1973), resulting in a similar radiation spectrum. However, in the general
case of a stochastic magnetic field, the radiation spectrum deviates strongly from this simpler
case.

It should be emphasized, that the applicability of the perturbation theory developed
above is itself rather limited. Indeed, even if the deflection of the particle is small during the
time needed to pass through a single correlation cell of the random field, it is not necessarily

(20)

(21)

4. Non-perturbative approach

- 8-

small along the coherence length of the emission, since multiple scattering of the particle by
several successive magnetic inhomogeneities can easily provide large enough deflections due
to angular diffusion to render perturbation theory inapplicable.

Thus, we explicitly have to consider the applicability of perturbation theory to the
emission of fast particles moving in the small-scale random fields. The coherence length
I c {uj) ~ c^l/u of emission by relativistic particle with the Lorentz-factor 7 at a frequency uj
decreases as frequency increases, so the approximation of the rectilinear motion is eventually
valid at sufficiently large frequencies. However, at lower frequencies the coherence length
may be larger than the correlation length of the random field, and this necessarily will be
the case at low enough frequencies if one neglects the effect of the wave dispersion in the
plasma. Thus, the particle trajectory traverses several correlation lengths of the random field
to emit this low frequency, and its trajectory random walks due to uncorrelated scattering by
successive magnetic inhomogeneities, as depicted in fig. 1. This angular diffusion will clearly
affect the radiation spectrum if the mean angle of the particle deflection 6 C accumulated
along the coherence length l c exceeds the beaming angle of the emission 7 -1 .

To estimate the deflection angle 9 C , adopt that the random field consists of cells with
the characteristic scale Z and rms field value (B^) 1 ^ 2 . Inside each cell, the electron velocity

1 /2

rotates by the angle 6q ~ (uu st /^)(lo/ c) ~ uj st l '(k'oT), where 00 st = e (Bg t ) /mc and uj =
2-kc/Iq. Since the deflections are produced by successive uncorrelated cells of the random
field, the mean square of the deflection angle after traversing N cells will be Q 2 C = O^N, where
the number of the cells is specified by the ratio of the coherence length l c to the correlation
length Iq of the random field, N = l c /lo- Therefore, when the particle passes the length
l c {u) required to produce the emission at a frequency u, its characteristic deflection angle is
Q\ ~ uo 2 st / (uoquo). This angular diffusion will strongly affect the emission if 9 C ^ I/7, i.e., at

cu < ^ 7 2 , (22)
^0

which occurs in the range of the low-frequency asymptotic limit discussed above.

The non-perturbative version of the theory is not discussed in any detail here since it
is published elsewhere (Toptygin & Fleishman 1987a,b; Toptygin et al. 1987; Fleishman
2006). The full radiation spectrum (also including the effect of a regular magnetic field, if
present) is given by Eq. (35) of (Toptygin & Fleishman 1987a). Fig. 2 presents radia-
tion spectra (calculated with the use of the non-perturbative formulae (35) of (Toptygin &
Fleishman 1987a), or, equivalently, Eq. (30) of (Fleishman 2006)) for a single relativistic
electron with a Lorentz-factor 7 = 10 4 moving in the presence of fluctuating small-scale fields
(with different ratios oo st /ujQ = 10 _1 , 10~ 2 , 10~ 3 , the regular magnetic field is assumed to be
very weak, the distribution of random field is isotropic (like (15)), and u pe = u st . Although

- 9-

the combination of parameters adopted to produce Fig. 2 does not span all the regimes
possibly relevant to the GRB case (see also the review paper Fleishman (2006)), it shows
the variety of asymptotes in one plot and displays spectra both affected and unaffected by
the non-perturbative effects, which should be sufficient for illustration purpose.

The solid curve corresponds to the parameters similar to those adopted by Medvedev
(2000) for the GRB sources (uj = 10u st = 10u pe ). Although the flat region obtained already
within the perturbation approach exists, it is easy to see a region of the radiation spectrum
dl w /duj oc cj 1 / 2 , provided by the multiple scattering, which occupies the frequency range

J 2

u pe (u pe cu -f 2 /cu 2 st ) 1/3 < cu < -^ 7 2 . (23)

u

Note, that the left bound of this inequality is below the limit uj pe ^ imposed by the effect of the
wave dispersion in the perturbative treatment. Stated another way, in the non-perturbative
regime the wave dispersion comes into play at lower frequencies than expected based on
the perturbative treatment. This region of the spectrum (missing within the perturbation
theory) exists for any relativistic particle with

7 * (24)

i.e., for 7 ^ 10 for the adopted parameters c^o — 10o; st = 10u pe , which turns out to be
important for most (if not all) available relativistic electrons in the GRB source.

On the contrary, for smaller-scale fields, dashed and dash-dotted curves, the angular
diffusion becomes less and less important, which justifies the applicability of the perturba-
tion theory for corresponding conditions. At even lower frequencies the spectrum falls as
dl^/dou oc uj 2 due to the effect of wave dispersion in the plasma. The high-frequency end
of the spectrum is described by the power-law dl^/du oc ou~ u , i.e., mimics the shape of the
assumed spatial correlation function. Both these asymptotes are present in the perturbative
version of the theory.

5. Implications for interpreting emission from GRBs

Typically, the GRB spectrum can be well fitted with a two-component Band model
(Band et al. 1993), which represents the observed spectrum as a sum of low-energy com-
ponent, N(E 1 ) oc E" exp(— E^/Eq), and high-energy component, N(E y ) oc E@ with a > (3.
The distribution of the low energy spectral indices a of GRBs (Preece et al. 2000), which
is a bell-shaped curve with the peak at a « — 1 with FWHM about 1, and reaches values
a m 1, is a challenge for available emission mechanisms (Baring & Braby 2004; Piran 2004).

-10-

Indeed, optically thin synchrotron emission can only produce low energy indices a <
—2/3 which are incompatible with about 25% of the spectra. Inverse Compton emission (as
well as a few other more exotic processes) seems to be more promising since it can produce
any negative spectral index a < 0, which is compatible with roughly 98% of the spectra. The
optically thick synchrotron model (Baring & Braby 2004; Lloyd & Petrosian 2000) might
accommodate the rest 2% of the positive a.

However, it is remarkable that none of these bounding values (a = —2/3, 0, 1.5) has
any imprint on the distribution: it is equally smooth at a ~ —2/3 or a ~ and does not
even reach values a ~ 1.5.

By comparison, the GRB low-energy-spectral-index distribution appears to be a natural
outcome of diffusive synchrotron emission. Leaving the detailed analysis and discussion of
some specific issues (Piran 2004) to a future study, we note that the main low-frequency
asymptote, oc u>°, corresponds to the peak of the distribution, which suggests plausible
interpretation why most of the low energy indices fall around the value a = — 1. Then, for
lower frequencies, this asymptote gradually gives way to the asymptote oc co 1 ^ 2 , related to
the multiple scattering effect, which is compatible with about 90% of the spectra. Therefore,
the part of the diffusive synchrotron radiation spectrum including these two asymptotes is
capable of explaining the whole range of the "typical" low energy spectral indices (which we
define as a confidence interval around the mean value). The remaining 10% of the spectra
might be treated as "untypical" , probably requiring some extreme combination of parameters
or some special structure of the random field. In any case, they are compatible with the
transition to the lowest-frequency asymptote, oc uo 2 . It is especially interesting that the latter
asymptote corresponds well to the presence of a secondary (weak but significant) peak in
the distribution at about a = 1.

In contrast, the high-frequency asymptote, oc w~ u , represents the softest possible high-
frequency spectrum in the presence of small-scale random magnetic fields. For example, if
v = 1.5, than the softest power spectral index is 1.5 and the softest photon spectral index is
2.5 independently of the relativistic electron distribution over energy (evidently, the spectrum
can be harder than this for hard enough electron distributions).

6. Summary and conclusions

Motion of a charged particle in the presence of random fields generally represents a
kind of spatial and angular diffusion, therefore, we refer to the emission by this particle as
diffusive synchrotron radiation. This paper presents a special case of the theory of diffusive

- 11 -

synchrotron radiation produced by relativistic electron deflections on a small-scale random
magnetic field in the absence of the regular magnetic field. This theory includes both the
simplified perturbation approach and the non-perturbative treatment, which turns out to be
of primary importance in the radiation spectrum formation.

The obtained spectrum of diffusive synchrotron radiation substantially expands the "jitter"
radiation spectrum suggested by Medvedev (2000) based on semi-quantitative evaluations.
The derived spectrum is composed of several power-law asymptotic regimes smoothly giving
way to each other as the emission frequency changes, which is apparently consistent with
the observed distribution of the low energy spectral indices in GRBs.

We conclude that

• diffusive synchrotron radiation has a very general nature and will be observed from
any source capable of producing small-scale random magnetic/electric fields at a sufficient
level, like active galactic nuclei, extragalactic jets, hot spots in the radio galaxies, galactic
sources with a strong energy release, as well as solar flares,

• a low-frequency spectrum, dl^jduj oc uj 1 , valid in the presence of ordered small-scale
magnetic field fluctuations, does not occur in the general case of small-scale random magnetic
field fluctuations,

• diffusive synchrotron radiation arising from the scattering of fast electrons on small-
scale random magnetic or/and electric fields produces a broad variety of low-frequency spec-
tral asymptotes - from dl w /du oc u° to oc uj 1 - sufficient to interpret the entire range of
low energy spectral indices observed from GRB sources, while the high-frequency spectrum
dl w jduj oc u~ v may affect the corresponding high energy spectral index distribution.

The National Radio Astronomy Observatory is a facility of the National Science Founda-
tion operated under cooperative agreement by Associated Universities, Inc. This work was
supported in part by the Russian Foundation for Basic Research, grants No. 03-02-17218,
04-02-39029. I am strongly grateful to T.S. Bastian for his numerous comments to the paper
and D.A. Frail for discussion of the topic.

REFERENCES

Band, D., Matteson, J., Ford, L. et al. 1993, ApJ, 413, 281
Baring, M.G., & Braby, M.L. 2004, ApJ, 613, 460

- 12 -

Epstein, R.I. 1973, ApJ, 183, 593

Epstein, R.I., & Petrosian, V. 1973, ApJ, 183, 611

Fleishman G.D., 2006, in "Geospace Electromagnetic Waves and Radiation", Eds. -
J.W.Labelle & R.A.Treumann, Chpt. 4, pp. 83-102, Lect. Notes in Phys., V.687
(Springer- Verlag: Berlin-Heidelberg-New York), preprint astro-ph/0510317

Jaroshek, C.H., Lesch, H., & Treumann, R.A. 2004, ApJ, 616, 1065

Jaroshek, C.H., Lesch, H., & Treumann, R.A. 2005, ApJ, 618, 822

Hededal, C.B., & Nishikawa, K.-I. 2005, ApJ, 623, L89

Kazimura, Y., Sakai, J.I., Neubert, T., & Bulanov, S.V. 1998, ApJ, 498, L183

Landau, L.D., & Lifshitz, E.M. 1971 The classical theory of fields (Oxford: Pergamon Press)

Lloyd, N.M., & Petrosian, V. 2000, ApJ, 543, 722

Medvedev, M. V. 2000, ApJ, 540, 704

Medvedev, M.V., & Loeb, A. 1999, ApJ, 526, 697

Meszaros, P. ARA&A, 40, 137 (2002)

Nikolaev, Iu.A., & Tsytovich, V.N. 1979, Phys. Scripta, 20, 665

Nishikawa, K.-I., Hardee, P., Richardson, G., Preece, R., Sol, H., & Fishman, G.J. 2003,
ApJ, 595, 555

Nishikawa, K.-I., Hardee, P., Richardson, G., Preece, R., Sol, H., & Fishman, G.J. 2005,
ApJ, 622, 927

Piran, T. 2004, Rev. Mod. Phys., 76, 1143

Preece, R.D., Briggs, M.S., Mallozzi, R.S., Pendleton, G.N., Paciesas, W.S., & Band, D.L.
2000, ApJS, 126, 19

Toptygin, I.N. 1985, Cosmic rays in interplanetary magnetic fields (Dordrecht, D. Reidel)

Toptygin, I.N., & Fleishman, G.D. 1987a, Ap&SS, 132, 213

Toptygin, I.N., & Fleishman, G.D. 1987b, Radiophys. & Qant. Electr., 30, 551

13

Toptygin, I.N., Fleishman, G.D., & Kleiner, D.V. 1987, Radiophys. & Quant. Electr., 30,
334

Weibel, E. S. 1959, Phys. Rev. Lett., 2, 83

This preprint was prepared with the AAS IATgX macros v5.2.

-14-

Fig. 1. — Random walk of the particle in the random field (a) vs jitter in small-scale regular field
(b). In the former case the mean square of the deflection angle 9 C increases with time as 9% oc t,
while in the later case it remains small along the entire trajectory. As a result, the radiation spectra
deviate strongly from each other for these two cases.

- 15 -

-i — ■
CO

CD

10° i
10" 1 i

10

; — 1 1 llllll

— 1 1 nun

— 1 1 nun

— i i nun

— i i nun

— 1 1 nun

c

— i i iiuii| — i i iiiiii

oVco .=

— 1 1 Mill

10

— 1 1 llllll

— 1 1 llllll

— 1 1 llllll

— i i iiiiii

1/2

CO

u

si

1

;"i

o 2 "

z

C
CO

*

/ /
/

\\

\

/ i
/ i

< /
/

*

-v !

co :

2 j
CO /

! i

*

f

CO t

St

= 1

CO

pe

V N

A i
\ V

iiiii 1 1 1 i |

/ /
/ /

' /
/

/ /

*

i

y=

10 4

i /

— 1 1 1 (ww

' i
i

— 1 1 Him

— 1 1 linn

— i i nun

— 1 1 nun

— i i iiiiii

— i i iiiiii

— i i iiiiii

— 1 1 IIIII

1 1 llllll

— 1 1 llllll

— i i iiiiii

— 1 1 llllll

10 10 10 10 10° 10

10

10

12

Frequency, co/co

pe

Fig. 2. — Radiation spectra produced by a relativistic particle with 7 = 10 4 in a small-scale random
magnetic field. The spectral asymptote w 1 / 2 can only be obtained within the non-perturbative
theory. If u>q is large enough (e.g., uoo/uj s t = 10 3 in the figure) the spectral region provided by
multiple scattering, u 1 / 2 , disappears, and the whole spectrum is the same as within the perturbation
theory.